MUMBAI, India, June 26 -- Intellectual Property India has published a patent application (202641073326 A) filed by Sr University on June 12, 2026, for Adaptive Deep Learning-Based System And Method For Real-Time Phishing Detection, Classification, And Prevention Of Cyber Threats.
Inventors include Dr. Arivukarasi M; and Dr Balajee Maram.
The application for the patent was published on June 19, 2026, under issue no. 25/2026.
Abstract: The present invention relates to the development of a system and methodology that utilize adaptive deep learning to detect, classify, and prevent cyber threats. This system uses an adaptive deep learning architecture that integrates convolutional neural networks for extracting features from the URL structure and content of emails and transformers to perform contextual linguistics analyses. The architecture has an online learning function that adapts the model's parameters to accommodate emerging phishing patterns. The system performs multi-class classification by categorizing legitimate content, credential harvest, malicious attachment delivery, and domain impersonation. It is designed to offer real-time operation via edge computing and thus, it provides less than 100ms latency for implementation in browsers and email gateways. The experiments conducted with the dataset of 2.3 million samples show that the system offers up to 98.7% accuracy, 97.2% precision, and 98.1% recall, which surpasses the performance of other rule-based and static machine learning methods. FIG.1
Disclaimer: Curated by HT Syndication.
